mirror of
https://github.com/SK8-ENGINE/skate-3-rust-engine.git
synced 2026-10-26 19:47:44 +00:00
Hails
4c043959c0
- AUDIO_TIMING logs voices / instances / grains per block (tools/audio-bench summary prints them). - The render path is allocation-free in real play: subscriber lists and constructor lists read in place, call parameters on the stack, instance memory and client lists from pools (render allocations per block 0.17-0.63 -> <= 0.001; full world 1.825 -> 0.004); render_alloc tests cover walks, posts, calls, globals and a retail world host. - Voices: direct PCM reads when a block provably stays inside the sample; shared scratch planes in the mixer (full world render p50 -11 to -14 %). - MixMap keeps its held-input list between ticks; Jitter processes without a per-frame Vec. - Every change byte-identical in all 84 e2e outputs; two attempted changes dropped (no real gain).
Audio bench
Tools for speeding up the native audio without changing what it does: measure first, prove the output
is identical, then compare timings. Built on the headless e2e render (see tools/audio-e2e).
| Script | What it does |
|---|---|
e2e_bench.sh |
Renders every scenario (and optional real-play sessions) headless with timing on, in one or more host modes (row: one call per 60 Hz row; fpsN: the game's host at N fps), and writes SHA-256 hashes of every output plus the timing log under a label. Uses its own cargo target folder so it never invalidates your main build. |
compare_runs.sh |
Compares two labels: output hashes (must be IDENTICAL) and the timing summaries side by side. Exit 1 on any difference. |
timing_summary.py |
Pools the per-frame (game thread) and per-block (render) times of one or more runs: p50 / p90 / p99 / p99.9 / max and total. |
audio_timing_summary.py |
Summarises the AUDIO_TIMING lines a real play session logs with SKATE_AUDIO_TIMING=1: per audio system, the average, the per-second maximum (median / p90 / worst) and calls per second. |
emitter_bank_memory.py |
Resident memory of the world emitter banks per .ems emitter file (or per group of files, e.g. a map's), from the install's audio manifest. |
prefetch_memory_sim.py |
Simulates a distance prefetch of emitter banks over a play session's audio state log and reports loaded and prefetched memory peaks. |
Inputs
BENCH_DIR(default.local/audio-bench):scen/*.tsvfromtools/audio-e2e/scenarios.py BENCH_DIR/scen, optionalreal/*.tsvfromscenarios.py --from-log LOG --cut r0-999999 NAME BENCH_DIR/real.- An installation with the native audio data (setup group
audio). - Game logs (
logs/game-*.stderr.log) from a session run withSKATE_AUDIO_TIMING=1. - Audio state logs from a session run with
SKATE_AUDIO_STATE_LOG=<path>.
Usage
bash tools/audio-bench/e2e_bench.sh base # before the change (keeps the renders)
bash tools/audio-bench/e2e_bench.sh new row fps30 fps300
bash tools/audio-bench/compare_runs.sh base new
py -3.13 tools/audio-bench/audio_timing_summary.py logs/game-20260101-120000.stderr.log
py -3.13 tools/audio-bench/emitter_bank_memory.py --group University=sfx_university,music_university
py -3.13 tools/audio-bench/prefetch_memory_sim.py my_state_log.tsv sfx_university --ahead 60 --evict 90
KEEP=1 keeps the .f32 renders of a non-base run; CARGO_TARGET_DIR overrides the bench target folder.
Example output
compare_runs.sh (made-up timings):
scen-row: IDENTICAL (26 files)
scen-fps300: IDENTICAL (26 files)
scen-row block:
base: n 41000 p50 210.0 p90 260.0 p99 340.0 p99.9 520.0 max 900.0 sum 8610.0 ms
new: n 41000 p50 130.0 p90 170.0 p99 230.0 p99.9 400.0 max 700.0 sum 5330.0 ms
Requirements
- Git Bash (
sha256sum,diff), the Rust toolchain. - Python 3.13 (standard library).
compare_runs.shcallspython(setPYTHONto change it).